Software Project Management
Towards Failure Avoidance
Thang N. Nguyen
Department of Information Systems, California State University Long Beach, California, U.S.A.
Keywords: Software Project, Management by Exception, Management Decision Evaluation, Biologically-inspired
Application.
Abstract: Development tasks, in the thousands or more, are involved in a complex system/software project. These
tasks aim at three objectives in the project, namely on budget, on time and on performance. Behind these
tasks are decisions made on them to move the project forward. Project failure is an aggregated and
cumulative failure of these tasks due to errors in tasks and/or decisions made. The project errors from faulty
strategies to wrong builds, collectively called exceptions are much harder to detect than cost overrun
(amounts spent) or time delay. The paper suggests a biologically-inspired approach to project failure
avoidance which is different than most. It focuses on exposing exceptions as they occur and understanding
the decisions made on them. Thus, there are two pieces needed for the proposed approach. The base piece is
a management by exception (MBE) framework to monitor development exceptions occurred for
management attention. The second piece is an adapted measurement method which will elicit, analyse and
evaluate the decisions made on the reported exceptions. We argue that using the proposed approach, failure
avoidance is possible and software project performance (in scope, intended features and desired quality)
would be under control, and so are expected cost and time.
1 ON SOFTWARE PROJECT
FAILURE
From the project perspective, according to Calleum
Consulting which combined various facts and
figures on why software fails (Calleum, 2014), it
was reported that “IT projects run 45 percent over
budget and 7 percent over time, while delivering 56
percent less value than predicted” (Calleam, 2014;
McKinsey and Co. survey), “Fuzzy business
objectives, out-of-sync stakeholders, and excessive
rework mean that 75% of project participants lack
confidence that their projects will succeed” and “a
truly stunning 78% of respondents reported that the
“Business is usually or always out of sync with
project requirements” (Calleam, 2014; Geneca
survey).
The KPMG survey showed “An incredible 70%
of organizations have suffered at least one project
failure in the prior 12 months and 50% of
respondents also indicated that their project failed
to consistently achieve what they set out to achieve
(Calleam, 2014; KPMG survey). In change
management, software projects experienced “40% of
projects met schedule, budget and quality goals, best
organizations are 10 times more successful than
worst organizations, biggest barriers to success
listed as people factors: changing mindsets and
attitudes – 58%, and corporate culture – 49%, lack
of senior management support – 32%,
underestimation of complexity listed as a factor in
35% of projects” (Calleam, 2014; IBM survey).
Other statistics (Mieritz, 2012) as well as
investigations on what happened, why they
happened, lessons learned etc. appeared in the
literature (Heusser, 2013; Galorath, 2011; Krigsman,
2008; Charette, 2005).
In this paper, we approach what make projects
fail differently i.e. from the perspectve of preventing
them. Our purpose is simply to detect exceptions,
whatever they are (monitoring issue), expose them
to appropriate parties (transparency issue) for proper
decision on them by the responsible parties, whoever
they are (control issue), and evaluate whether the
decisions are arbitrary, risky or otherwise
(justification issue) towards failure avoidance.
560
Nguyen T..
Software Project Management - Towards Failure Avoidance.
DOI: 10.5220/0004992605600567
In Proceedings of the 9th International Conference on Software Engineering and Applications (ICSOFT-EA-2014), pages 560-567
ISBN: 978-989-758-036-9
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
In a complex system/software project, there exist
thousands or more tasks. These tasks aim at three
objectives in a project, namely on budget, on time
and on performance (in this paper, on performance
implies: in-scope, intended features or
functionalities, and desired quality). Behind these
tasks are decisions made on them by decision
makers (DMs), i.e. management and subject matter
experts (SMEs), exercising some decision making
schemes.
In terms of success-failure, with pure monitoring
of the funding amounts spent, task start dates and
completion dates, the cost and time of the software
project will reflect cost overrun or time delay for
corrective decision making. Software performance,
however, is more complex to monitor and measure
in every phase of development. The project errors
from faulty strategies to wrong codes, hereafter
collectively called exceptions are commonly much
harder to detect until they occur. These, if not
detected and/or if occurred but not addressed
properly for any reasons, could aggregate and
accumulate into more critical wrongdoings
(exceptions) and would bring the project to failure.
By and large, the success or failure of any tasks
can be attributed to the decisions made by the
management (including top executives) and software
developers (including other SMEs such as business
analysts, etc.). They are the responsible parties. They
are the people who execute one task to the next
and/or sign off the specifications, documents or
other artifacts.
We propose to monitor the exceptions during the
life of the project, which might be resulted from
decisions, arbitrary or otherwise. Since all project
tasks are linked in a complex decision network in a
critical path-like, the overall failure can be initiated
by the first wrongdoing initiated by some decision
contributing to the final failure of the software
project. We attempt to understand and to measure
these decisions on exceptions made by the
management-leadership team and the developers-
SMEs responsible for the project.
Thus, there are two major pieces of our proposed
solution. The base piece is a event-driven framework
which will house a amangement by exceptions
(MBE) application to monitor and to expose all
development exceptions occurred during the
development cycle for management and developers
attention. The MBE is responsible by an Oversight
organization (or Committee). This organization can
then requests management-SMEs-developers to look
at the decisions they make on the exceptions.
Thus, the second piece is a measurement method
which elicits, analyses and evaluates the decisions
made on the reported exceptions and consequences.
Applying this measurement scheme for timely
correction, we argue that we could possibly, at some
confidence level, avoid failure, and software
performance would be at the expected level in terms
of scope, intended features and desired quality, and
so are cost and time.
2 A FRAMEWORK FOR
ENTERPRISE-WIDE
INFORMATION
MANAGEMENT BY
EXCEPTION
For small size projects with fewer managers and
developers, an MBE application can be simply daily
or weekly meetings where exceptions are reported.
For mid-size project, a software development
decision model for managing software development
projects as suggested by (Nguyen, 2006) or any
project management tool in the market could be
appropriate. For larger or very large project such as
(1) the now-defunct Future Combat System of the
US Army in the 2000’s (GAO, 2009), (2) the US Air
Force Expeditionary Combat Support System
ECSS, and (3) the US Marines Global Combat
Support System - GCSS (GAO, 2012; Kanaracus,
2012), or the recent difficulties experienced by
Healthcare.gov (Schadler, 2013), we would ask if a
different scheme is possible for detecting exceptions
and measuring decisions made on them.
To that end, we exploit a couple of
considerations towards a framework for an
information management by exception which is
biologically-inspired. We discuss the rationale of
such framework in some details in this section.
The initial consideration stems from the
biological spectrum (Figure 1) which consists of
protoplasm component at the lowest end to the
biosphere component at the highest end (Alberts et
al., 1998; Raven, 2008).
Some part of this biological spectrum has been
the source for insights by different researchers
during the last century. At the cell level, for
example, there have been the Computer and The
Brain and the Theory of automata by John Von
Neumann (Von Neumann, 1966), and the theory of
autopoesis (Maturala and Varella, 1980), to name
just a few. Institution (as community) and business
SoftwareProjectManagement-TowardsFailureAvoidance
561
Figure 1: The biological spectrum.
ecosystems in the sense of James Moore (Moore,
1996) are part of the spectrum.
We look at the spectrum a little differently,
however. We wonder if both human (as organism)
and institution (as community) can be viewed
analogously (the blue box with red text in Figure 1).
That is if an institution is considered as analogous to
a human body (structural, functional or behavioral)
then the employees of the institution are analogous
to the cells of the human body.
From an exception perspective, cells in the
human body can turn abnormal or cancerous. If the
abnormal cells grow uncontrollably, invade nearby
tissues, termed as malignant tumor, and
subsequently proliferate to other organs, the tumor
can bring death to the human (King, 1996). Note that
an infectious disease by a deadly virus in a host cell
would also cause death.
Analogously, if a group of people in an
institution turned abnormal for any reasons
(commonly greed, power, growth, risk, etc.), they
can become an “institution malignant tumor”. If this
group is funded and exercises their influence to
other organizational units, they can bring collapse to
the institutions. Examples of cancer-like
wrongdoings which led to collapses in the financial
circle are Enron, WorldCom, Adelphia, Parmelat,
and Lehman Brothers (Foster, 2010).
In software development projects circle, issues
leading to exceptions such as ill-understood business
problems, risky contracts with client, out-of-sync
between stakeholders, underestimated complexity of
solutions, and others (Pressman, 2010) are
originated from top management-SMEs-developers’
decisions. These decisions, if arbitrary or turned bad,
when aggregated, might bring failure to the project.
We can extend the analogy between human
(organism) and institution (community) to software
product (these are products of humans in the
institution) into perspective as shown in Figure 2.
The analogy does not have to be perfect as long as it
can offer some insights into the making of our
proposed solution, subject to verification and
validation.
For humans, at the higher level there are
biological guiding principles that govern the
structure, functionality and behavior of a human
body at the middle level. They are: (1) the “milieu
interior” of Claude Bernard (Gross, 1998) in which
all cells, tissues, organs and organ systems of the
body reside, (2) the principle of cybernetics (Wiener,
1948) which controls the human functionality and
behavior, and (3) a condition called homeostasis
where the human body maintains its equilibrium
(Cannon, 1963).
Figure 2: Human-Institution-Software analogy.
The corresponding analogous principles of the
institution, considered as a community are namely,
information environment, managerial cybernetics
(Beer, 1972) and stability as shown in the middle
column. At the lower end, we recognize the five
elements of the human body supporting the
structure, function and behavior of the human.
Analogously, we list the corresponding analogous
elements supporting an institution. Those of
software products are shown on the right side of
Figure 2.
Figure 2 will not help much in our attempt to
modelling, unless we re-arrange all the guiding
principles at the top level, the structural, functional
and behavioural aspect at the middle level, and all
the supporting elements at the lowest level of Figure
2 to reflect activities and events that occur as shown
in the dotted box of Figure 3 and in the control
mechanisms.
ICSOFT-EA2014-9thInternationalConferenceonSoftwareEngineeringandApplications
562
Figure 3: Biological framework.
Figure 4: Business framework analogous to biological
framework.
From this rearrangement, we can formulate two
other frameworks which are considered analogous,
to some extent, to the biological framework (Figure
3). They are: the business framework (Figure 4) and
the development framework (Figure 5). Note that in
Figure 3, the set of proteins, macromolecules,
cellular exchanges and DNA/genes are analogous to
those in Figure 4, with projects, tasks, transactions,
accounts and policy, and in Figure 5, with objects,
classes, program calls, components, and program
language references. All three frameworks share a
common goal: the detection of exceptions.
The monitoring and detection in the last two
frameworks (business and development) is
performed by an MBE application as exemplified in
the application menu shown in Figure 6.
The MBE captures software project data, e.g. at
the management level, it involves Project, Tasks,
Transactions, Accounts, and Policy (e.g. System
Operating Procedures or high-level strategy), and at
the supporting development level it involves objects,
classes, program calls, components, and standard
operating procedure (SOP) or programming
language references, with the intention to detect
exceptions much like the detection of malignant
tumor in the human body.
Figure 5: Software project framework analogous to
biological framework.
Figure 6: MBE application menu.
The data acquired (Data Acquisition on the left of
Figure 6) are subject to different analysis methods
listed in the middle part of Figure 6 (Analysis).
Different exception reports (Reporting) can be
requested and produced.
Note that we don’t have to draw the cancer
analogy to arrive at an MBE for software projects as
proposed. However, the cancer analogy gives
insights into the criticality of symptoms of a
malignant tumor. Cancerous symptoms are quite
often hidden until later stages (by the time they
surface, it is too late, the human would die). This can
be parallel to the hidden wrongdoings in software
project failures or in corporate collapses. The
analogy gives another advantage: it is possible that
some known biological processes would give
insights into the formulation of additional analysis
schemes or methods since it is the humans who
create institutions and software products, or any
other man-made products.
The three umbrella frameworks depicted by
Figure 3-5 might have some weaknesses. They
might need some fine tuning to explore the details of
the analogical aspect. But for our purpose, they are
sufficiently adequate. They show commonalities in
terms of the activities, events and control
mechanisms guided by policy for the detection of
exceptions to be exposed to management as sketched
in Figure 6. The exceptions are considered as
symptoms of wrongdoings on which decisions by
SoftwareProjectManagement-TowardsFailureAvoidance
563
the responsible parties are made.
It is known that the responsible parties do not
always properly act on exceptions as expected. At
times, decision makers might intentionally involve
in arbitrary decisions. They can also purposely
ignore, avoid, alter or hide the symptoms which
could aggravate or lead to more catastrophic
situation. Therefore exposing symptoms of
wrongdoings and making them transparent are
necessary but not sufficient. There must be a way for
the Oversight organization to “force” the responsible
parties to take actions and measure their decision’s
effectiveness. This is the topic of the next section.
3 MEASURING DECISION
EFFECTIVENESS
Decisions made by Good management-leadership
and Good employees-SMEs would ensure success
(quadrant 2 of Figure 7). This is contrasted and
opposed to Bad management-leadership and Bad
employee-SMEs (quadrant 4). The latter would
possibly bring to project failure. If one of the two is
bad while the other one is good, it will be more
complex to measure or label the failure-success
(quadrant 1 or 3) of the decisions.
Figure 7: Management-Leadership top grid.
In all four cases (quadrants) we wish to know
how to provide a measure of the decision maker’s
effectiveness on a scale of failure-success in the
software development management. The top grid of
Figure 7 actually embodies underneath a hierarchical
decision grid network which can be very complex.
In this grid, we think of employees-SMEs as
primarily concerned with operational and some
tactical decisions in an institution. On the other
hand, management-leadership is primarily involved
with strategic and some tactical decisions, and
corporate vision.
Thus, within the context of this top grid of
Figure 7, we need to drill-down any exceptions in
software development projects, on which decision
makers made decisions.
Example decisions which might cause
exceptions are: A sales person closes a development
contract with minimum involvement of technical
personnel, leading to under-estimated size and
complexity. A business analyst insists on unrealistic
requirements. A developer manager approves the
use of open-source for cost reasons. Example
exceptions include: A tester reports a memory leak.
A client experiences a deadlock between multiple
users using the same input form against the same
policy or rule on data inputted. From this set of
exceptions, we need to elaborate all possible
attributes of the decisions on exceptions, which we
will call constructs.
The proposed process to arrive at constructs is
adapted from George Kelly’s Personal Construct
Theory (PCT) (Kelly, 1963), Valerie Stewart’s
repertory grid (RG) (Stewart, 1981, 2010) and other
variations (Smith et al., 1996).
We ask the decision makers to qualify the
exceptions as business (and/or technical) elements
(projects, tasks, transactions, accounts, and policy)
associated with the exceptions. They have to select
three exceptions at a time (called a triad), identified
as crucial in terms of decision made on them. We
will ask:
In what way two of the elements have in
common, (emergent construct) and
In what way both of them differ from the third
(implicit construct, opposite to emerging)
We build a detailed grid with columns headings
as the exception elements identified by triad, and
with rows as the hierarchical structure of decision
constructs in opposing pairs underneath the top
grid. Hinkle’s laddering up and Landfield’s
laddering down (Fromm, 2004) and other recent
modifications to laddering are used to elicit other
constructs (Korenini, 2012).
We ask them to associate each decision construct
with a ranking or better yet a rating. Specifically,
they will be asked to rate each and every construct in
a 5-point (or 7-point scale), e.g. from 1-5, with 1
being at one end of the construct dipole (emergent)
and 5 being its opposite (implicit). Content analysis
and/or cluster analysis are used as exemplified in
(Stewart, 1996) and (Bourne and Jenkins, 2005), and
others.
The purpose is to identify the significant
decision constructs involved and their rated values
for all elements. From the grid, a final and combined
measure of all evaluations expresses a level of
success (or failure) in the overall failure-success
ICSOFT-EA2014-9thInternationalConferenceonSoftwareEngineeringandApplications
564
dipole. This is the indicator of how effective the
management-leadership and SMEs-developers team
member is.
The elicitation, analysis and evaluation is
responsible by a separate organization called an
Oversight organization or Committee. The skills set
for elicitation, analysis and evaluation needs be
developed.
4 DISCUSSION
In section 2, we proposed a biologically-inspired
framework (event-driven, Figure 5) and the MBE
application (Figure 6) for the detection and
exposition of exceptions in an institution by the
human-institution analogy. One would expect that
when examining, in section 3, the decisions on the
exceptions we would look into neuroscience
(Kandel, 2000) or cognitive neuroscience (Koch and
Davis, 2003) for analogy insights into the
institution’s decisions made by its management-
leadership-SMEs team.
The bottom-up path (anatomy-physiology-
neuroscience) however is too complex for us to
handle and offers no guarantee that we can reach the
top grid constructs for success-failure (Figure 7)
since decision topics in neuroscience is still under
investigation, even with the best know technique,
fMRI. Instead, we chose to address decision on
exceptions top-down, from a psychological aspect in
the sense of George Kelly. This is investigated with
the hope that the top-down approach to decision
would lead to a construction system which would
identify some core constructs eventually delineated
by cognitive neuroscience findings. This is to
establish an integration link between the exceptions
(low-level events) and the decisions on them (high-
level action). Thus, the repertory grid in section 3
helps reveal the what, why and how (in what way) of
a hierarchical decision construct model by which the
responsible parties in an institution, individually and
collectively, respond to different exception
situations.
Our thinking is that if we can discover and
understand the construction system of the decision
makers (by elicitation as discussed in section 3) on
exceptions in this particular domain: software
development and management, then not only we can
(1) explain what caused software failure, (2) predict
the consequences, (3) take remedy actions, but also
(4) avoid future crises.
There are two scopes in pursuing the top-down
approach. First, by performing the elicitation
individually, the Oversight organization begins to
pay attention to the DM responsible or involved in
the exception for recommended remedy action and
prediction measure to avoid future crisis or failure.
Secondly, by analysing the elicited data collectively,
some patterns characterizing the institution’s
decision model can be discovered.
Since the repertory grid technique has been
investigated over the last five decades and have been
applied to many domains successfully, even though
it is complex and requires excellent interview skills,
we can be assured at some level of confidence that
the technique would work in our domain of interest:
software development and management.
Thus, the main issue is not the power of the
repertory technique but a thorough understanding of
the domain so that we could adapt the technique
successfully and effectively.
So, one of the relevant questions is “what is
involved in this domain, and how can we apply the
RP technique?” The software development, like any
other development projects is creativity-driven. The
main component is termed as peopleware (Demarco
and Lister, 2012). The average person are highly
trained and well experienced. But as in any situation,
there exist two other groups: the outstanding versus
the weak, the quality-driven versus the error-prone,
the greedy versus self-content, the quick versus the
slow, the abuser versus the abused, etc. all represent
the bipolar concept on constructs. What we finally
construe is a collection of constructs revealed by
individuals which can organized collectively. We
would be able to characterize the institution as a
whole, within the context of the four quadrants of
the top grid (Figure 7).
There are a couple of issues (not exhaustive) on
the construct system under investigation.
Selection of triads: One would think that since
the elicitation is based on a specific exception that
occurs during the life of the project, how would the
decision maker (DM) select a triad of exceptions to
work with, let alone many different triads? First,
note that the DM is not anybody in the institution.
The DM is either a top management member, or a
SME with a wide range of responsibilities. The DM
influences the success-failure of the project, since
the DM heavily is involved in the project in many
managerial, business and technical aspects of the
project: planning, scheduling, funding, personnel,
technology, training, support, skills, etc. The DM
should and would be familiar with many prior
exceptions during his tenure with the institution or
elsewhere. So, there won’t be problems that the
DMs identify the other elements to work with the
SoftwareProjectManagement-TowardsFailureAvoidance
565
exception at hand.
The interviewer: According to many RG authors,
the interviewer must be skilful. The interviewer
can’t suggest any construct. The interviewer’s
questions are based primarily on the DM’s responses
to further explore the DM’s decision model in terms
of laddering up or laddering down. The interviewer
must faithfully scribe the information as it is given.
Rating the grid: If 1-5 scale rating is used instead
of the two-value scheme suggested by George Kelly,
is it possible that the DM fakes the rating? On a
scale of 1-5, 3 being neutral or not applicable (N/A),
4 and 5 are normally assigned to the emerging
construct, with 1 and 2 assigned to the implicit
construct. The faking of rating, if occurred, will not
impact the final analysis a great deal.
Analysis and evaluation: This is the task of the
Oversight organization. Analysis can be done using
a commercially available tool. Content analysis or
statistical tools such as cluster analysis can be used
to evaluate similarity and difference among the set
of decisions on exceptions. The expected result from
this analysis and evaluation offers an understanding
on the consensus or otherwise among DMs on the
issues, circumstances, actions, consequences, and
values, and other factors surrounding the exceptions.
We would see whether the decision is arbitrary or
not, or whether decision makers are high-risk driven,
etc. Again, the grid also can expose individual
perception and thoughts of the decision maker on the
exceptions. Thus, it could help discover any
improper intention. That’s the basis of our proposed
psychological method adapted from Kelly and
others.
5 CONCLUDING REMARKS
Our biologically-inspired investigation is driven by
insights into the biological spectrum. On the surface,
it appears that it is similar to previous investigations
in the past in terms of insights and/or metaphors. For
example, one could say that the Object-Oriented
(OO) inheritance concept is drawn from Gregor
Mendel (Alberts, 1998) who discovered the principle
of inheritance, basis of genetics. The window icon
was after Charles Peirce’s theory of signs (Stanford
2006). Many OO design patterns were after
Christopher Alexander’s creation in the field of
architecture (Alexander, 2002).
However, our approach is different in that the
biological spectrum as a whole offers a global and
integral scope. In the software project management
domain, we can extent further to include exceptions
caused by suppliers (e.g. primary or subcontractors)
and other stakeholders in an IT development project
(such as (1) the case of US Army Future Combat
systems, (2) the US Air force ERP project which
could have been pulled out sooner, rather than
waited until after $1B already spent, or (3) the
Marine Corps' GCSS project which could have
avoided delay and budget overrun (Kanaracus,
2012).
We can include medical providers, insurers,
patients, etc. such as in the case of Healthcare.gov,
(Heusser, 2013) where each group is considered as
population or community, and the whole system of
systems as a business ecosystem. Interesting insights
can be gained from this perspective. It allows the
concept that humans (as organism) assembled in
institutions (as a community) which run business as
part of (business) ecosystems within some economy
(biosphere).
Furthermore we couple the biologically-inspired
framework with a psychologically-driven technique
on decisions to characterize decisions, because we
recognize that a wrongdoing, if not properly
handled, can lead to other wrongdoings of higher
criticality, and therefore aggravates the project
health. This is worse especially when the decision
makers try to hide their bad decisions in the process
for one reason or another.
In the financial world such as the Barings Bank
case, Nicolas Leeson was able to hide the loss of his
first trading transaction between Osaka exchange
and Singapore exchange (SIMEX) in the error
account 88888 (Leeson, 2012), without management
knowledge (maybe his Singaporean subordinates
knew but did not report). He was also responsible
for both the front office as a trader and the back
office as a general manager of Barings Singapore
(an oversight by his managers and executives in
London Office). Both should be exposed earlier as
exceptions for proper decisions, and the bank could
have avoided collapse.
In the case of Enron, CFO Andrew Fastow, was
able to offset losses in the Enron financial
statements over many years using a complex
structure of Special Purpose Entities (SPE) as
hedging scheme (Powers, 2002). If the first warning
(Mack in 1993) to Enron CEO Ken Lay, or if the
following warnings reported by Enron’s own
accountants on the use of SMEs were not overridden
or actually received proper attention, additional
wrongdoings by Andrew Fastow and his team would
have been avoided (Powers, 2002).
Other solutions such as prevention of corporate
fiascos, can be formulated by looking more closely
ICSOFT-EA2014-9thInternationalConferenceonSoftwareEngineeringandApplications
566
at the biological spectrum as we have discussed in
section 2 and exemplified in the case of Barings
Bank and Enron bankruptcy, and also as found in
(Nguyen, 2014).
REFERENCES
Alexander, Christopher, 2002. The Nature of Order, Pub.
The Center for Environmental Structure.
Alberts, Bruce, Dennis Bray, Alexander Johnson, Julian
Lewis, Martin Raff, Keith Roberts and Peter Walter.
(1998). Essential Cell Biology, Garland Publishing.
Beer, A. Stafford,1972. Brain of the firm: Managerial
Cybernetics of Organization, The Penguin Press,
London 1972.
Bourne, Humphrey and Mark Jenkins, 2005. Eliciting
Managers' Personal Values: An Adaptation of the
Laddering Interview Method, Organizational
Research Methods 2005 8: 410
Calleam, 2014. Why Projects Fail: Facts and Figures,
Calleum Consulting. .http://calleam.com/WTPF/?
page_id=1445
Cannon, Walter, 1963. The wisdom of the body, The
Norton Library, Norton & Company, 1963
Charette, Robert N., 2005. Why Software Fails? IEEE
Spectrum. Volume 52, Issue 9. September 2005.
Demarco, Tom and Timothy Lister, 2012. Peopleware:
Productive Projects and Teams, http://javat
roopers.com/Peopleware.html.
Foster School of Business, Washington, 2010. A decade’s
worst financial scandals, April 2010
Fransella, Fay and Bannister, Don, 1977. A Manual for
Repertory Grid Technique, Academic Press, 1977.
Fromm, Martin, 2004. Introduction to the Repertory Grid
Interview, Waxmann Verlag.
GAO, 2009. Review of Future Combat System Is Critical
to Program's Direction, http://www.gao.gov/new.
items/d08638t.pdf.
GAO, 2012. 12-565R: DOD Financial Management
http://gao.gov/assets/590/589796.pdf
Galorath, Dan, 2012. Software Project Failure Costs
Billions. http://www.galorath.com/wp/software-proje
ct-failure-costs-billions-better-estimation-planning-
can-help.php June 2012
Gross, C. G., 1998. "Claude Bernard and the constancy of
the internal environment" Neuroscientist 4 (1)
Heusser, Mathew, 2013. 6 Software Developments
Lessons from Healthcare.gov’s Failed Launch, http://
www.cio.com/article/743366/6_Software_Developme
nt_Lessons_From_Healthcare.gov_s_Failed_Launch
Kanaracus, Chris, 2012. The Scariest Software Project
Horror Stories of 2012. http://www.cio.com/
article/print/721628
Kandel, Eric and James Schwartz, Thomas Jessell, 2000.
Principle of Neuroscience, 4
th
edition, (Eds), McGraw-
Hill, 2000
Kelly, G., 1963. A Theory of Personality: The Psychology
of Personal Constructs. New York: W. W. Norton &
Company
King, Roger J.B., 1996. Cancer Biology, Longman, 1996.
Christof, Koch and Davis, Joel L., 2003. Large-Scale
Neuronal Theories of the Brain, Christof Koch and
Joel L. Davis (Eds), MIT Press
Korenini, Bojan, 2012. Conducting Consisting Laddering
Interviews Uisng CLAD, Metodološki zvezki, Vol. 9.
No.2.
Krigsman, Michael, 2008. 68% of IT Project Fail, Beyond
IT Failure, December 2008, http://www.zdnet.com/
blog/projectfailures/study-68-percent-of-it-projects-
fail/1175
Leeson, N., 2012. Rogue Trader, Little Brown Book Group.
Mack, T., 1993. Hidden Risks, Forbes.
Maturana, Humberto R. and Francisco J. Varela., 1980.
Autopoesis and Cognition: The Realization of the
Living, D. Reidal Publishing Company.
Mieritz, Lars, 2012. Gartner Survey Shows Why Project
Fails? June 2012, http://thisiswhatgoodlookslike.com/
2012/06/10/gartner-survey-shows-why-projects-fail/
Moore, J.F., 1996. The Death of Competition – Leadership
and Strategy in the Age of Business Ecosystems,
Harper Business, 297.
Nguyen, T. N., 2014. A Different Approach to
Information Management by Exceptions: Towards the
Prevention of another Enron, Information &
Management, Vol 51 Issue 1.
Nguyen, T. N., 2006. A Decision Model for Managing
Software Development Projects, Information &
Management, Vol 43 Issue 1.
Pressman, Roger S., 2010. Software Engineering: A
Practitioner’s Approach. 7e. McGraw-Hill.
Powers Jr., W. Report of Investigation, 2002.
Stewart, Valerie., 2010. http://valeriestewart-repertory
grid.blogspot.com/
Raven, Peter H. George B. Johnson, Jonathan Losos,
Susan Singer, 2008. Biology, 8e, McGraw-Hill
Reilly, Sean, 2012. How the Air Force blew $1B in a dud
system. Federal Times, http://www.federaltimes.
com/article/20121126/DEPARTMENTS01/31126000
9/How-Air-Force-blew-1-billion-dud-system
Schadler, Ted, 2013. Healthcare.Gov’s Failure Starts with
Leadership, Not Technology, http://www.forbes.com/
sites/forrester/2013/10/25/healthcare-govs-failure-
starts-with-leadership-not-technology/
Stanford Encyclopaedia of Philosophy, 2006. http://
plato.stanford.edu/entries/peirce-semiotics/
Stewart, Valerie, Andrew Stewart with Nickie Fonda,
1981. Business Application of Repertory Grid,
McGraw-Hill Books Co.
von Neumann, 2000. The Computer and The Brain, 2e,
Yale University Press.
von Neumann, John; Burks, Arthur W., 1966. Theory of
Self-Reproducing Automata
, University of Illinois
Press.
Weiner, Norbert, 1948. Cybernetics or Control and
Communication in the animal and the machine, The
Technology Press.
SoftwareProjectManagement-TowardsFailureAvoidance
567